Oh wise Perl magicians

I have a query that looks something like this :

select ra.value, ra.ra_id, q.id as q_id, q.qid , q.parent, q.q, q.type
+, p.qid as pqid, if($ifstatement,'1','0') as selection , r.rpid from 
+r, ra, q left join q p on q.parent = p.id where r.id = ra.ra_id and r
+.q in (3,4,5) and ra.q = q.id order by ra.ra_id ASC,selection desc;
[download]

The 'explain' shows a good and nice run (no full table scans), and in mySQL it takes about 3 seconds to run (about 130.000 records) using  mysql -u user -p database < q.sql > output.txt
When doing this from whithin Perl however,

my $bm=  DEBUG_Benchmark->new();
$bm->bmStart();

$sth = $dbh->prepare("select ra.value, ra.ra_id, q.id as q_id, q.qid ,
+ q.parent, q.q, q.type, p.qid as pqid, if($ifstatement,'1','0') as se
+lection , r.rpid from r, ra, q left join q p on q.parent = p.id where
+ r.id = ra.ra_id and r.q in (3,4,5) and ra.q = q.id order by ra.ra_id
+ ASC,selection desc");

$sth->execute();

$bm->bmStop();
print $bm->getBmResult();
[download]

It takes about 12 seconds to run (Only the prepare and the execute, i did not even loop over the resultset).
Any idea on how to speed this up? Whit a smaller resultset, it takes about .003 seconds to prepare / execute.
I know DBI should be able to do mutch better, any ideas on how i can get a better result?
Thanks.

ps.
I run this on an UML (user mode linux), so there is a lot of overhead in the disk io, however, the same problem occurs on a 'normal' debian distribution, the difference is big. (2 sec - 8 sec)

ps.ps.
The resultset of the query looks like this: *** UPDATE ***
Ah FOUND IT!!! Here is the answer!!!